Towards a Usage-Based Cognitive Phonology 1

34
© Servicio de Publicaciones. Universidad de Murcia. All rights reserved. IJES, vol. 6 (2), 2006, pp. 107-140 IJES UNIVERSITY OF MURCIA International Journal of English Studies www.um.es/ijes Towards a Usage-Based Cognitive Phonology 1 GITTE KRISTIANSEN * Universidad Complutense de Madrid ABSTRACT The usage-based conception of language is a major tenet in Cognitive Linguistics, but cognitive phonology has not yet been developed sufficiently in this direction. Often, phonemic analysis is carried out at the high level of abstraction of `a language´, disregarding rich patterns of language- internal variation. This paper first argues that cognitive phonology must aim at a higher degree of descriptive refinement, especially in the direction of social variation. Then it goes on to examine the implications of a usage-based and multi-faceted model for a theoretical discussion of the phoneme as a prototype category. KEYWORDS: usage-based cognitive phonology, lectal variation, distributed cognition, receptive and active competence, language acquisition, models of phonemic representation * Address for correspondence: Gitte Kristiansen. Universidad Complutense de Madrid. Facultad de Filología. Departamento de Filología Inglesa I. Lengua y Lingüística. 28040 Madrid. Phone: +34913945392. Fax: +34 913945478. e-mail: [email protected]

Transcript of Towards a Usage-Based Cognitive Phonology 1

© Servicio de Publicaciones. Universidad de Murcia. All rights reserved. IJES, vol. 6 (2), 2006, pp. 107-140

IJES

UNIVERSITY OF MURCIA

International Journal of

English Studies www.um.es/ijes

Towards a Usage-Based Cognitive Phonology1

GITTE KRISTIANSEN* Universidad Complutense de Madrid

ABSTRACT The usage-based conception of language is a major tenet in Cognitive Linguistics, but cognitive phonology has not yet been developed sufficiently in this direction. Often, phonemic analysis is carried out at the high level of abstraction of `a language´, disregarding rich patterns of language-internal variation. This paper first argues that cognitive phonology must aim at a higher degree of descriptive refinement, especially in the direction of social variation. Then it goes on to examine the implications of a usage-based and multi-faceted model for a theoretical discussion of the phoneme as a prototype category.

KEYWORDS: usage-based cognitive phonology, lectal variation, distributed cognition, receptive and active competence, language acquisition, models of phonemic representation

* Address for correspondence: Gitte Kristiansen. Universidad Complutense de Madrid. Facultad de Filología. Departamento de Filología Inglesa I. Lengua y Lingüística. 28040 Madrid. Phone: +34913945392. Fax: +34 913945478. e-mail: [email protected]

Gitte Kristiansen

© Servicio de Publicaciones. Universidad de Murcia. All rights reserved. IJES, vol. 6 (2), 2006, pp. 107-140

108

I. ON THE NECESSITY OF A USAGE-BASED FRAMEWORK FOR COGNITIVE PHONOLOGY

A recent Google search on `accent reduction´ (January 7, 2007) gave 2.040.000 results. Most of

the first 400 results corresponded to courses offered by private companies or public institutions

(universities included) aiming at a reduction of foreign accents. However, a surprisingly high

number also had as their target the reduction of native accents. In such cases, the course

descriptions often explain in quite straightforward terms that having a regional accent can `keep

you from being promoted´ or `hamper you professionally and socially´. It is easy to find

companies that specialize in `American Regional Dialect Reduction´ and offer courses or

products which teach the unfortunate speakers with regionalisms such as `American South and

Texas´, `Mid-West farm Belt´, `American Urban and Rural Black´ or even `New York City and

North Jersey´ how to `speak without an accent´. It has thus not gone unnoticed that there is

market for teaching people how to speak with socially prestigious accents. The unspoken

assumption is that accents are socially diagnostic: humans have `passive´ competence of lectal

varieties (cf. Kristiansen, 2003) in the sense that we possess the ability to process clusters of

linguistics cues which quickly and efficiently signal social and regional origin (hearer categorizes

speaker lectally and socially) and concomitantly invoke the corresponding social stereotypes

(hearer characterizes speaker socially and geographically).

One of the most important cues to correct dialect identification (perhaps even the most

significant one: cf. Purnell et al., 1999; van Bezooijen & Gooskens, 1999) is allophonic variation.

However, while the disciplines of Sociolinguistics and Social Psychology of Language have

explored the relationship between linguistic variants and social meaning for more than four

decades now, social and regional variation has strangely enough been relegated to a secondary

position in cognitive phonology. Yet, there are good reasons for investigating these areas in more

depth within a Cognitive Linguistics framework. For a start, it would seem more than just

pertinent, the emphasis on meaning-making processes provided, to investigate not only the

mechanisms by means of which allophonic variation evokes social meaning, but also the various

options language users employ once the link between linguistic form and social domain has been

established (cf. Kristiansen forthcoming a). Further, an analysis of linguistic variants in actual

Towards a Usage-Based Cognitive Phonology

© Servicio de Publicaciones. Universidad de Murcia. All rights reserved. IJES, vol. 6 (2), 2006, pp. 107-140

109

usage within social and linguistic dimensions which taxonomically speaking are more specific

than the large-scale and abstract category of `a language´ is only in line with the fundamental

claim that Cognitive Linguistics is a usage-based perspective. As Langacker, the scholar who

coined the term, once phrased it:

In a usage-based model, substantial importance is given to the actual use of the linguistic system and a speaker’s knowledge of this use; […] It is a non-reductive approach to linguistic structure that employs fully articulated schematic networks and emphasizes the importance of low-level schemas (Langacker, 1999: 91).

In turn, Geeraerts (2001, 2005; Geeraerts et al., 1994) has repeatedly drawn attention towards the

logical entailment of such a position: we can only take the claim that Cognitive Linguistics is a

usage-based approach seriously if the kind of language that we analyze is real language, language

as it is actually used by real speakers in real situations. Obviously, an analysis which in a natural

way incorporates social factors and other types of language-internal dimensions not only provides

us with a far more realistic picture, but also widens the gap with respect to disciplines according

to which it is still acceptable, and possible, to analyze languages in terms of idealized speakers

and homogeneous speech communities.2 It naturally follows that a fine-grained map of

allophonic variation within a given speech community cannot ignore social variation of the type

just described.3 Lectal variation will naturally form part of any description, or model, which

purports to provide a realistic picture of phonetic variation and phonemic categorization.

However, the implications are not only descriptive, but also theoretical. From a diachronic

perspective, knowledge about the social significance of linguistic variants may well turn out to

have an influence on active competence, and accordingly on the nature of the variants in actual

usage – both quantitatively and qualitatively - in a given language at a given historical time. A

theory of language that intends to describe and explain the dynamics of language change in

adequate ways cannot afford to ignore the (synchronic) mechanisms of actual language usage.

Obviously, to obtain a picture which allows us to discuss the nature and evolution of not only

phoneme categories but also phoneme inventories, the very first step is to aim at a high level of

Gitte Kristiansen

© Servicio de Publicaciones. Universidad de Murcia. All rights reserved. IJES, vol. 6 (2), 2006, pp. 107-140

110

descriptive accuracy. Usage-based models, then, take the data as they actually appear and set up

theories which conform to the facts.

As far as the link between usage and language acquisition is concerned, Taylor (2002: 27)

summarizes the relationship in the following way:

It is assumed that the input to language acquisition are encounters with actual linguistic expressions, fully specified in their phonological, semantic, and symbolic aspects. Knowledge of a language is based in knowledge of actual usage and of generalizations made over usage events. Language acquisition is therefore a bottom-up process, driven by linguistic experience.

Though I shall also touch on acquisition in what follows, my main concern will be the descriptive

and theoretical implications of zooming in on real usage in a more persistent way.

For the sake of highlighting the importance of social variation, I have started out with this

particular aspect, but a usage-based model is obviously not one-dimensional. In fact, if Cognitive

Linguistics shares an interest in low-level schemas situated at the level of parole with

Sociolinguistics, with Functionalism it shares, amongst many other aspects (cf. Nuyts, 2005) the

view that meaning cannot be studied in isolation, separable from the nature and the purpose of

what is communicated and from the dynamics of communication as such. In the next section I

focus on three of the dimensions which have a bearing on allophonic variation, viz. the

ideational, discursive, and social functions of language. In other words, I will discuss (a) the kind

of phonetic variation which serves the fundamental purpose of realizing a distinctive unit (i.e. the

phoneme) and which does not necessarily carry an additional social message, (b) co-textual

variation (that which derives from discursive factors), and (c) variation of a contextual type (that

which pertains to social cognition and interaction).

II. A MULTI-DIMENSIONAL APPROACH TO PHONETIC VARIATION

In what follows I shall discuss a number of different dimensions implementing the terms

`ideational´, `social´ and `discursive´. These correspond, roughly speaking, to Halliday’s (1978)

trichotomy of the `ideational´, `interpersonal´, and `textual´ functions of language. However, on

Towards a Usage-Based Cognitive Phonology

© Servicio de Publicaciones. Universidad de Murcia. All rights reserved. IJES, vol. 6 (2), 2006, pp. 107-140

111

the assumption that not only inter-personal, but also inter-group and intra-group mechanisms are

at work in social interaction, the term `social´ will be employed instead of `interpersonal´.

Likewise, `discursive´ is preferred to `textual´, since, although we know that texts – in the sense

of coherent stretches of language regardless of size or mode – are oral as well as written, the

dynamics of ongoing, oral, discourse is made more prominent by using the former expression. I

retain, albeit reluctantly, the term `ideational´ since social information is no less a question of

(internally complex and subjectively construed) semantic domains than is `factual´ information.

In this respect, Jakobson’s (1960) `referential´ and Lyon’s (1977) `descriptive´ do not offer a

helpful distinction, either.

II.1. The ideational function of language

The ideational function of language, according to which phonemes function as builders of

`factual´ meaning at the level of the morpheme and allophonic variants are processed as mere

phonemic slot-fillers, is so well-known that little needs to be explicated. In very general terms,

since the discovery of minimal pairs, both Structuralism and Functionalism established a clear

distinction between the meaning-making levels of linguistic structure (beginning with

morphology) and the meaning-building ones (i.e. the levels `below´ morphology). It was not until

the birth of Sociolinguistics that the correlation between social variables and phonetic variants

began to be studied in a systematic way. Cognitive Linguistics has perhaps neglected precisely

that aspect, but in turn it has contributed to a prototype-theoretical conception of the phoneme as

a mental category (Mompeán, 2004; Nathan, 1986, 1994, 1996, 1999; Taylor, 1990, 1995, 2002).

In a phonemic prototype category, phonetic variants cluster around a central member which is

maximally contrastive with respect to the central members of neighboring categories (Taylor,

1995: 221). In theory, then, the specific sets of phoneme categories organized around such major

acoustic-perceptual contrasts in actual usage (the options logically being constrained by human

physiological conditions in general) constitute the phoneme inventories of natural languages.

However, although the cognitive phonology view of phoneme category structure is

certainly an attractive model, it is far from unproblematic. For a start, it is implicitly assumed that

Gitte Kristiansen

© Servicio de Publicaciones. Universidad de Murcia. All rights reserved. IJES, vol. 6 (2), 2006, pp. 107-140

112

the vast kind of language-internal variation we encounter across lectal categories can be brought

under one schematic representation. It is assumed, for example, that the English phoneme

category /t/ can be described (implementing either a network model or a radial category model) in

such a way that all the variants of /t/ in actual usage in `English´ can be subsumed within the

same category. In the case of /t/, even if social variants are allowed to become reflected, we can

still easily talk about a consistent radial category, featuring extensions based on relative similarity

with respect to central or more peripheral members, all of which ultimately organized, in more or

less direct ways, around a prototypical member. Yet in many other cases, the model is not that

easily applied. Or rather, viewed from a usage-based perspective, the data fail to fit the model in

as neat a way as we would like them to. This is particularly true in the case of the vowels. Let us

take an example from the Linguistic Atlas of England (Orton et al., 1978):

Figure 1. Variants of the vowel in the verb lay in traditional English dialects according to The Linguistic Atlas of

England (Orton et al., 1978).

Towards a Usage-Based Cognitive Phonology

© Servicio de Publicaciones. Universidad de Murcia. All rights reserved. IJES, vol. 6 (2), 2006, pp. 107-140

113

The map in Figure 1 represents realizations of the vowel pronounced in the word-form <lay> in

traditional dialects in England between 1950 and 1960. The data are perhaps somehow obsolete,

but they will do for our present purposes, as the heterogeneous situation conveyed by the map is

representative of dialectal variation. We may note, for a start, that many of the steps of the Great

Vowel Shift never reached the very North of England, with retention of many of the

monophthongs from the Old English period as a result. In fact, while it is certainly still possible

to draw up a schematic representation which depicts chaining relationships among the variants, or

instantiations, in actual occurrence, it is much harder to posit that there should be just one single

phoneme (not in the sense of a prototype category, but of a distinctive builder of ideational

meaning) at work at the same time. We encounter realizations such as [laɪ] and [li:], which would

clearly be transphonemic if one specific phonological system, such as RP, were to form the basis

of our model (i.e. [laɪ] and [li:] evoke the semantic poles of lie and lee for a speaker of RP and

many Southern English dialects). The question is not only how to chart the internal structure of a

formal category in adequate ways, but also whose system(s) we are representing. Large-scale

speech communities are complex and heterogeneous, and to be really usage-based cognitive

phonology must account for the fact that there are multiple, and quite dissimilar variants in use

and different phonological systems at work within the same language. The easy way out is to

adopt the position that each lectal variety forms an autonomous system of its own, to be analyzed

independently, regardless of the existence of neighboring, or adjacent systems. As I have pointed

out on other occasions (cf. Kristiansen, 2003: 76), this is precisely the way systemic-functional

linguistics solved the problem. By establishing a distinction between language as institution

(consisting of independently formed varieties) and language as system (language perceived as a

system analyzable in terms of layers of linguistic structure), Halliday (1978) sifted one fairly

homogeneous (but of course still very heterogeneous in many other respects) kind of language

from the sum of its lectal varieties. Yet a usage-based approach cannot afford to work at such a

high level of abstraction, especially in the field of cognitive phonology. It is not realistic to work

around one variety of a language, no matter how prestigious and accessible it happens to be, and

equate it with the language in question.

Gitte Kristiansen

© Servicio de Publicaciones. Universidad de Murcia. All rights reserved. IJES, vol. 6 (2), 2006, pp. 107-140

114

II.2. The discursive function of language

Natural phonology argues that the discursive roles of hearer and speaker result in a series of

(perhaps conflicting, but on the other hand perfectly compatible) tendencies. While both hearer

and speaker seek improvements, speaker does so by means of mechanisms involving `ease of

effort´ and hence `ease of articulation´. The hearer-oriented role, in turn, is rather aimed towards

`clear and effective communication´. Following Stampe (1979), Nathan (1996: 116-117) has

argued that fortitions are processes which select among all possible human sounds those which

constitute the phonemes of a particular language and which define prototype effects for language

sounds. Lenitions, on the other hand, are processes which create allophonic variants, i.e.

extensions from a prototypical member within a radial network. These phonological processes are

“universal cognitive mechanisms that languages have and may or may not use (that is, may or

may not suppress) in any given instance” (Nathan, 1996: 113). Fortitions thus help us understand

phoneme inventories in terms of series of phoneme categories whose prototypical members show

a maximum degree of perceptual difference. Such an approach is of course highly compatible

with general principles of prototypicality: maximum acoustic-perceptual salience and inter-

categorial contrast as an organizing principle for the structure of phoneme inventories and

categories constitute criterial factors in both cases. Numerous processes of assimilation, on the

other hand, may be explained in terms of lenitions (e.g. the `rule´ according to which an alveolar

nasal becomes labialized in immediate contact with other labial sounds: <in Paris> [ɪmˈpærɪs]).

Obviously, the discursive (or co-textual) variants which systematically occur in a given dialect -

or at least in the most prototypical instantiations of the variety in question – would naturally form

part of a minute description of intra-phonemic variation.4

II.3. The social function of language

Accents are socially diagnostic. When a stretch of speech is processed, not only can hearer

decode an ideational message (conveyed by constructions from those layers of a language we

traditionally label in terms of phonology, morphology, syntax or lexis), but also an additional

social message. This can be even more forceful than an explicit statement, precisely because

Towards a Usage-Based Cognitive Phonology

© Servicio de Publicaciones. Universidad de Murcia. All rights reserved. IJES, vol. 6 (2), 2006, pp. 107-140

115

hearer on occasions receives the information in an implicit way. The information we receive

without being fully aware of it is likely not to be questioned in the same way as an explicit

statement about regional origin and psychological characteristics would. Paradigmatic variation

thus overrides the linear constraint of language in subtle ways. To the extent that children

gradually build up knowledge of lectal varieties and learn how to relate them systematically to

social domains, social values and stereotypical perceptions, a speaker can become not only

categorized, but also characterized on the mere basis of his accent.

If this (relative) awareness has implications for language change, it may ultimately also

have an influence on the shape of phonemic categories and inventories. In this respect, when a

given phonetic variant begins to spread throughout the social and regional dimensions, ceases at a

given point, co-exists with other variants and eventually replaces a number of its local

`competitors´, the role and the motivations of the speakers who, either above or below the level

of conscious awareness, opt for just this variant should not be underestimated. In fact, as Bybee

points out, in order to understand language change, we should look for dynamic mechanisms

which pertain to or influence actual language usage:

…the true universals of language are the dynamic mechanisms that cause language to change in certain systematic ways as it is used and as it is transmitted to new generations. (Bybee, 2001: 189)

The kind of social cognition that will be described in more detail in this section can be viewed to

fall within the category of such dynamic mechanisms. In a multi-dimensional approach to

language variation and change, cognitive, social and functional factors will naturally combine as

causative mechanisms to eventually shape the systems of a given language. This is a conception

which is ultimately not compatible with theories which regard social factors as mere triggers of

deeper, inherent or `natural´ causes of language change (e.g. Aitchison, 1991).

At this stage it is necessary to clarify that in this paper the term `social´ is used to denote

the various social contexts which surround us. On the one hand there is an immediate context

which involves the speech event as such, including the participants (speaker, hearer, bystanders,

etc.) and their social status, the setting, the topics, the communicative goals, etc.). On the other

Gitte Kristiansen

© Servicio de Publicaciones. Universidad de Murcia. All rights reserved. IJES, vol. 6 (2), 2006, pp. 107-140

116

hand, there is also a wider social context which involves the participants’ knowledge of, or belief

in, a series of the Cultural Cognitive Models (Holland & Quinn 1987) at their disposal, and their

knowledge about social groups, including social and linguistic stereotypes. These two types of

context intertwine and interact in intricate ways. In fact, they are only distinguished in such an

apparently easy and discrete way here for the sake of explanatory clarity. The former situates

social interaction in real time and space and enables us to work around actual (and often

purposive) usage of speech styles. The latter is transmitted, negotiated, developed and maintained

in situated social interaction. Both types of context thus play a crucial role in social cognition.

In very general terms, in the author’s main line of research (e.g. Kristiansen, 2001, 2003,

forthcoming a) the fact that accents are socially diagnostic has served as a starting-point from

which related issues have been explored in a number of different directions, including the

relationship between accents (in terms of structured speech patterns) and social meaning from a

Cognitive Linguistics perspective. My research thus falls within the wider fields of cognitive

dialectology, cognitive sociolinguistics and language variation and change, but part of the

analysis has a direct bearing on phonology and may be summarized as follows:

1. Lectal varieties (i.e. those categories we more traditionally label in terms of regional or social

dialects, accents or speech styles) and social categories (i.e. social groups and identities such

as British, Cockney, Northerner, South African, Australian, etc.) constitute prototype

categories which interact at various levels of abstraction. The central images of lectal

varieties (speech templates or linguistic stereotypes, consisting of a cluster of salient features,

of which allophonic variants play an important role) act as effective reference point

constructions which through a basic metonymic operation (accents form part of a wider

frame, a social domain) evokes the corresponding social stereotypes: LANGUAGE STANDS FOR

SOCIAL IDENTITIES.

2. Once the link between a linguistic and a social stereotype has been established, it may be put

to even more constructive uses, as speakers posses not only receptive, but also – at least to

some extent – productive competence of speech styles. It is extremely difficult to imitate a

Towards a Usage-Based Cognitive Phonology

© Servicio de Publicaciones. Universidad de Murcia. All rights reserved. IJES, vol. 6 (2), 2006, pp. 107-140

117

non-native speech style to perfection, but the most salient features are relatively easy to

perform. Paradigmatic variation is also a metaphor: LANGUAGE IS A TOOL FOR CONVEYING

SOCIAL MEANING.

These are statements which need to be spelled out in more detail. When I implement the notion of

linguistic stereotype it is in a neutral, technical way. What I have in mind is a complex cluster of

features which, from the perspective of folk perception, in the best and clearest way allows us to

categorize and identify the structured speech style of the members of a given speech community.

Linguistic stereotypes very effectively evoke the corresponding social stereotypes, conceived, in

equally neutral terms, as outgroup images of a given social category.

Speech patterns acquired in early childhood are not easily changed, and linguistic

stereotypes thus constitute an especially reliable marker of social identity. Hence, the link

between linguistic and social stereotypes is fundamentally of a metonymic nature: an EFFECT FOR

CAUSE mapping which leads the conceptualizer from a linguistic trigger to a wider social target:

to a social group and the encyclopaedic knowledge we have about it (social habits, dress, dance,

song). This knowledge includes a series of Cultural Cognitive Models (i.e. ideological patterns

and components) and often take the form of stereotypical perceptions.

I understand social stereotypes in terms of simplified outgroup perceptions which condense

information regarding what the members of a given group are like (e.g. in terms of psychological

attributes, ideological beliefs and social behavior). The existence of a fairly stable relationship

between speech styles and social targets thus underlies the general metonymy LANGUAGE STANDS

FOR SOCIAL IDENTITIES. When a linguistic feature is heard as `prestigious´, `intelligent´ or `posh´

and in reality it is the group of speakers associated with the feature in question which is being

evaluated as such, it is accordingly not a process of iconization (Irvine & Gal, 2000), but rather

an indexical process which is at work. The force of the process is easily comprehended if we bear

in mind that the features contained in just one two-syllable word (cf. Purnell et al., 1999) suffice

to evoke the whole lectal category (a part-whole metonymy) which in turn links, again

indexically, with a social domain and the corresponding social stereotypes.

Gitte Kristiansen

© Servicio de Publicaciones. Universidad de Murcia. All rights reserved. IJES, vol. 6 (2), 2006, pp. 107-140

118

Receptive competence is however only the starting-point of a much more complex story.

Once a stable referential link has been established between a lectal variety and a social domain,

entrenched linguistic trigger - social value relationships can now be put to new, constructive uses.

In other words, the existence of receptive competence can be exploited by speakers in order to

signal social values. Structured paradigmatic variation can thus also be put to metaphorical uses:

LANGUAGE IS A TOOL FOR CONVEYING SOCIAL MEANING.

If I speak of the central images of lectal varieties and social groups as cognitive reference

point constructions (and not just cognitive reference points), it is to emphasize the fact that such

images are relative and relational: situated and group-dependent construals. Social and linguistic

stereotypes are perhaps best understood as instances of situated cognition (cf. Kristiansen,

forthcoming b), as structure which emerges, is transmitted, negotiated and maintained through

dynamic interaction between people in real situations in real historical time. In consequence, the

specific combination of items which compose a given social stereotype will to a large extent

depend on contextually determined intergroup relationships. In the case of linguistic stereotypes,

the features that characterize the speech of an outgroup will also be determined by the nature of

the features present in our own ingroup speech pattern. For instance, the phonetic variants which

are perceived as salient and identifying of French to an Englishman or a Dane might not be

viewed as such by an Italian. If a speaker possesses a similar realization in similar phonetic

contexts in his own language, the feature will not stand out as perceptually salient. Lectal

categorization, when viewed from this broad perspective, is not only a purely cognitive

phenomenon, but also a process which must also be considered in terms of cultural situatedness.

For linguistic stereotypes to relate in an exclusive way to a social group – for accents to be

socially diagnostic – the cluster of features that set a speech style off as distinct from others must

be composed of a unique combination of perceptually salient features. This is in consonance with

Nunberg’s (1978) line of thought when he asked himself how hearer and speaker manage to

determine referents in deferred ostension. In metonymic conceptual operations, what kind of

visual demonstratum will successfully lead to the intended referent, and which will fail to do so?

As I have previously pointed out (Kristiansen, 2003), Nunberg in reality asked himself what the

ideal signifier in operations of metonymic reference is like. He reasoned that a given form will

Towards a Usage-Based Cognitive Phonology

© Servicio de Publicaciones. Universidad de Murcia. All rights reserved. IJES, vol. 6 (2), 2006, pp. 107-140

119

successfully lead to the intended referent if it relates to it in an `exclusive´ way (so as to identify

and not just characterize the target) and that the form in question must be `perceptually distinct´

enough for effective subclassification to take place, so as to be able to distinguish it from similar

forms within the same general category which have different values attached to them.

In the field of phonology, there are multiple possibilities (though each language only

exploits relatively few of these) of establishing acoustic-perceptual contrasts which allow for

salient subsets to become effected within the more general category of a phoneme. There is no

reason why such minor contrasts should not function according to the same fundamental

principles as those which determine degrees of prototypicality for phoneme categories in general:

…the putative central member of /t/ -say, the voiceless aspirated alveolar plosive- enters into a number of highly salient perceptual and articulatory contrasts with the putative central members of neighbouring categories, such as the unaspirated alveolar plosive of /d/, the voiceless aspirated velar plosive of /k/, and so on. (Taylor, 1995: 228)

Intraphonemic acoustic-perceptual contrasts (subphonemic prototypes) are thus minor when

compared to that which sets [tʰ] off from, say, [kʰ], but still major enough to create distinct

subsets within a given category:

Figure 2. Some linguistic stereotypes based on intraphonemic acoustic-perceptual contrasts within the category /t/ in British English and metonymic reference to social domains.

Gitte Kristiansen

© Servicio de Publicaciones. Universidad de Murcia. All rights reserved. IJES, vol. 6 (2), 2006, pp. 107-140

120

As a case in point, the use of the glottal stop in intervocalic position is obviously not sufficient to

invoke the Cockney accent. Rather, it is a complex cluster of features (cf. Kristiansen, 2003,

forthcoming a) which from a hearer-oriented perspective effectively categorizes a stretch of

speech as a token of a given type in an exclusive way. From a speaker-oriented perspective, we

would speak in terms of social differentiation being achieved by means of linguistic

distinctiveness (cf. Giles et al., 1987; Tajfel & Turner, 1979).

Assimilation to a prototype category is usually thought of in terms of relative similarity with

respect to another member of the category in question, be this peripheral or more central. Hence

we speak of chaining relationships or radial networks. Prototype categories are flexible entities in

the sense that the boundaries are extendable (new members may be added to allow for human

cognition to adapt itself to change, innovation or new discoveries in a complex social and

physical world). In theory, a new member is similar enough to at least one existing member in

order to be categorized as a member of a given category and not as a member of a contrasting

category (or an instance of a given schema, not another). Conversely, however, a new member

must also necessarily be perceived as distinct, or different enough to deserve the status of a

different subcategorization, a new extension or a new instantiation, to use several of the notions

in current usage, and not just a token of an already existing type.

For the sake of exemplification, consider an invented case of categorization from the visual

domain. In Figure 3, the left-most shape is a kiki and the right-hand shape is a booba.5 Or so at

least 95 per cent of subjects systematically estimated in a series of experiments on psycho-

acoustics (Köhler, 1929, 1947; Werner, 1934, 1957; Werner & Wapner, 1952) when asked which

shape was called what in a language unknown to them, the options being kiki and booba (the

latter word pronounced with [ɔ:], not [u:]):

Figure 3. A kiki and a booba.

Towards a Usage-Based Cognitive Phonology

© Servicio de Publicaciones. Universidad de Murcia. All rights reserved. IJES, vol. 6 (2), 2006, pp. 107-140

121

Suppose that, as a reader of this paper, you are now asked to draw another kiki and that you draw

a shape which is very similar to the kiki represented above. Suppose that you are told that you

were expected to reflect a `different´ kiki, not the same kind of kiki (i.e. not a token of the same

type). You then draw a kiki with fewer or perhaps more sides than the first one, but presumably

still with pointed sides, so as not to produce a booba. If asked to do the same with the booba

shape, your drawing might resemble this one:

Figure 4. A different kiki and a different booba.

Say more instances of irregular kikis, or six-sided kikis, become produced, either in different

situations or by different groups of subjects. And that four-sided boobas show up systematically

alongside seven-sided boobas under a series of contextual circumstances. The general categories

of kikis and boobas have now become divided into subcategories, with slightly different tokens

representing two different subtypes. At the linguistic end of the formal trigger-conceptual content

axis, a new term would in all probability now arise to designate the new intra-categorical subtype,

and this form would in many cases iconically convey both membership (retain part of the word-

form designating the general category) and subclassification (possess some kind of formally

distinctive element). The adjectival modifier in `four-sided kiki´ serves the latter purpose, the

modified head the former.6 The point is of course that successful subcategorization seems to be

based as much on subtle, but still perceptually salient enough differences as on perceived

similarity. In a similar way certain allophones, those which I have referred to in terms of

subphonemic prototypes, while perceived as members of the same general category, are also

perceptually distinct enough to form a new subcategory and thus serve as ideal triggers of new,

additional meaning. Salient allophones can serve the dual purpose of realizing a phoneme

(according to the ideational function) and evoke social group membership at the same time.7

When viewed from both a hearer and speaker-oriented perspective, the possibility of drawing on

such a pool of triggers of social meaning enables hearer to decode social information on the one

Gitte Kristiansen

© Servicio de Publicaciones. Universidad de Murcia. All rights reserved. IJES, vol. 6 (2), 2006, pp. 107-140

122

hand, and allows groups of speakers to encode it by choosing especially contrastive forms in their

speech, much in the same way as they would, more mundanely, wear a distinctive kind of

garment or opt for a specific hairstyle.

III. LECTAL VARIETIES AS EXPERIENTIALLY GROUNDED CONSTRUALS

In this section lectal categories will be examined in terms of construals grounded in individual

and group-related experience. In III.1, I discuss the distinction between receptive and productive

competence of lectal categories and relate competence to the notion of relative awareness.

Subsection III.2 centers on lectal competence and language acquisition, and finally, in III.3 I

address the question of lectal competence and distributed cognition.

III.1. Receptive and productive competence

Languages are schematic with respect to their instantiations: we inevitably speak a given variety

of our mother tongue. In much the same way, linguistic input necessarily consists of real

instantiations which become processed for a variety of purposes in terms of low-level or high-

level schemas. For example, a phoneme is a schematic abstraction which cannot be pronounced

as such. Also, words are invariably realized by means of a combination of specific phonetic

variants. A word such as <butter> can only be realized as e.g. [ˈbʌtʰə], [ˈbʊtʰə] or [ˈbʌʔə] and then

processed `ideationally´ as a sequence of phonemes (/ˈbʌtə/ → `butter´). In the right

circumstances (i.e. when hearer possesses the necessary knowledge and is attentive enough) such

instantiations will also be processed lectally (a given combination of linguistic features leads us

to a particular lectal variety). That phonetic detail should be stored and processed alongside the

function of realizing a phoneme is certainly not at odds with Bybee’s (1988, 2001) model of a

mental lexicon and a usage-based phonology; words are stored in their concrete phonetic forms

and phonetic detail retained in long-term memory.

Let us assume that we gradually acquire knowledge about a large number of lectal varieties

and the speech communities they relate to. In other words, that we gradually acquire receptive

Towards a Usage-Based Cognitive Phonology

© Servicio de Publicaciones. Universidad de Murcia. All rights reserved. IJES, vol. 6 (2), 2006, pp. 107-140

123

competence of speech styles. Lectal categorization would then involve a conceptualizer who

correlates a token (stretch of unidentified speech) with a number of idealized speech models

(linguistic stereotypes). The similarity may be relative, of course; two people who `speak with the

same accent´ obviously do not speak exactly the same way. Rather, their intonation patterns,

phonetic realizations and phonemic slots are judged to be relatively similar when compared to a

model. But we also soon learn how to put linguistic stereotypes to other, equally constructive

uses. The fact that effective categorization seems to be based on a reduced series of highly salient

features facilitates the process known as style-shifting. While it is extremely difficult, if not

impossible, to imitate a non-native accent to perfection, with all the subtle phonotactic and

distributional combinations of salient and much less salient variants, the components of a

linguistic stereotype are fairly easy to imitate. The term productive competence denotes the use of

features from a style which does not form part of a speaker’s habitual repertoire.

A note is now necessary on awareness. Accents are socially diagnostic because linguistic

cues index social meaning, or – in more technical terms – because a source-in-target metonymic

conceptual operation mediates between structured sets of linguistic triggers and the social domain

they project. However, this process is presumably often a below-the-level-of-consciousness

affair. A little more than four decades ago, with the birth of Sociolinguistics, it became clear to

many scholars that a systematic study of social dialects could not rely on the same methods as

those traditionally implemented in the study of regional dialects. Eliciting informants’ intuitions

in a direct way, e.g. by means of the questionnaires used in many surveys, proved to be an

inadequate procedure for a systematic description of social dialects for two major and interrelated

reasons. On the one hand, social speech styles relate to contextual factors. Speakers vary their

style according to situational factors such as setting and topic, or the style and status of the

interviewer.8 They also evaluate, consciously or not, the relative position of their own style on a

social hierarchy of varieties according to variables such as prestige and stigmatization – and

might accordingly over-represent the actual occurrence of features which rate high on the scale of

prestige. On the other hand, actual usage, as the sociolinguists soon recognized, is often situated

below the level of conscious awareness. In consequence, speakers’ own perception of their

speech style is likely not to coincide with actual usage, self-perception being potentially

Gitte Kristiansen

© Servicio de Publicaciones. Universidad de Murcia. All rights reserved. IJES, vol. 6 (2), 2006, pp. 107-140

124

distorted, inaccurate or – if we take the principle of the cognitive unconscious seriously – quite

simply not fully accessible.9 Awareness is furthermore a gradable dimension, and speakers are

presumably often only conscious to a certain degree of the messages they receive when listening

to stretches of speech. However, messages received below the level of conscious awareness are

still received – and may lead to positive or negative evaluations, to rejection, admiration or

imitation, if only on an apparently intuitive basis. Finally, and with regards to productive

competence, that a speaker should occasionally imitate – or try to adopt on a regular basis – a

feature from a given speech style because it is `fashionable´ is a commonly quoted explanation of

sound change. Yet fashion can surely also be viewed as a variable which is ultimately dependent

on hearer’s perception of the hierarchical position of the lectal and social domains projected by a

given linguistic feature. This perception invariably involves at least an implicit degree of

awareness regarding the link between linguistic feature, social group and social meaning. Fashion

could thus also be viewed as a cover term for a limited, but still productive degree of awareness

as regards the processes which relate linguistic form to social meaning.

We all soon acquire a natural stylistic repertoire and moreover posses the ability to imitate

new speech styles – or at least the most salient features of the styles of other groups. We may

even set up new, local identities. These are often effected by selecting a series of socially

meaningful features stemming from stable, large-scale social categories. Eckert (2004), for

instance, reports on the various ways in which one particular British English feature (final /t/ with

an audible release of aspiration), associated with the British as superior, intelligent and educated,

has been put to a variety of different uses by American speakers of English (cf. Kristiansen,

forthcoming a).10

The features that are imitated, or adopted and put to new uses, in the first place are those

that stand out as especially salient. In this respect, not all contrasts are equally important in terms

of acoustic-perceptual perception, and this is one more reason why we should treat the social

function as separate dimension. It could be argued, for instance, that an account of phonetic

variation which incorporates the variants in actual usage within the lectal varieties that compose a

given language (alongside those that arise from discursive factors and other relevant functions)

will already include the kind of social variation which is under scrutiny in this section, i.e. that by

Towards a Usage-Based Cognitive Phonology

© Servicio de Publicaciones. Universidad de Murcia. All rights reserved. IJES, vol. 6 (2), 2006, pp. 107-140

125

lowering the level of abstraction so as to work around language-internal varieties and not `a

language´, the social function is already duly covered. But if we did that, we would on the one

hand remain at a descriptive level and fail to appreciate numerous factors which might well have

a bearing on language change, and on the other hand miss out on the opportunity to investigate

such factors in more depth. Language acquisition, for instance, is one of the areas in which it

might be particularly fruitful to invest.

III.2. Lectal competence and language acquisition

So far research on accent-based speaker identification (e.g. Purnell et al., 1999; van Bezooijen &

Gooskens, 1999; the many studies on speech identification in relation to Artificial Intelligence)

and perceptual dialectology (e.g. Niedzielski & Preston, 2000) has largely concentrated on adult

informants. Yet, if we adopt a usage-based approach and assume that the acquisition of speech

styles is experientially grounded and that phonetic detail is stored as such (Bybee, 2001) and put

to constructive uses, then there is a lack of empirical studies on the acquisition of receptive and

productive competence of lectal varieties in children. When do children begin to construe low-

level schemas, paying attention not only to what is said, but also to how it is said? If schemas are

usage-based, at what age does dialect identification emerge, and how specific is it at different

intervals of age? There is a need in other words to investigate, amongst others, the following

factors:

• The degree to which children acquire receptive competence of accents at different intervals

of age. • The relative precision with which accents are identified at different intervals of age. • The relative degree of awareness regarding the specific features which allow children to

proceed to correct dialect identification. • The relative capacity of children to imitate accents (productive competence) at different

intervals of age. • The factors which, age apart, have a bearing on lectal acquisition. • The relative degree of awareness of children regarding the relationship between linguistic

and social stereotypes.

Gitte Kristiansen

© Servicio de Publicaciones. Universidad de Murcia. All rights reserved. IJES, vol. 6 (2), 2006, pp. 107-140

126

III.3. Lectal competence and distributed cognition

If “input to language acquisition are encounters with actual linguistic expressions, fully specified

in their phonological, semantic and symbolic aspects” (Taylor, 2002: 27) and “knowledge of a

language is based in knowledge of actual usage and of generalizations made over usage events”

(ibid), it follows that not all individuals will possess the same kind of knowledge, nor always

effect the same kind of generalizations. Or as Geeraerts (1997: 110) formulates it, a distinction

between “a cumulative, macro-level picture of the language and the individual language user’s

micro-level knowledge” is most convenient:

[…] we would probably not want to maintain that all mature speakers of the language actively command the entire range of semasiological possibilities that are combined in the prototype-based descriptions. An alternative way to interpret diagrams […] is to think of them as representing the summed knowledge of language users at a certain moment in the development of the language (and then also, of course, the knowledge of an ideal language user). (ibid.)

From the perspective of a level of granularity which lies above that of the individual, Sharifian

(2003) argues that the elements of cultural schemas are not shared by all members of a cultural

network, but rather distributed across the minds. It is not by virtue of the belief in only one

schema that one becomes a member of a cultural group, but the overall degree of how much a

person draws on various cultural schemas that makes an individual a more or less representative

member. Cultural schemas, or cultural cognitive models, thus thrive within groups and the group

emerges as such, shaped and brought into existence by relatively shared beliefs, values and

norms. In similar ways, the group also determines and is determined by relatively shared speech

patterns such as `dialects´, `accents´ and `styles´ and by relatively shared social stereotypes.

In this respect it would be interesting to know more about how uniform folk perception of

the way in which another social group speaks is across the members of a given ingroup. We

might also want to know more about the extent to which linguistic stereotypes constitute relative

construals across different cultural and lectal communities. To what extent, for instance, does the

linguistic stereotype of Spanish differ when acquired by a Frenchman, an Italian and an

Towards a Usage-Based Cognitive Phonology

© Servicio de Publicaciones. Universidad de Murcia. All rights reserved. IJES, vol. 6 (2), 2006, pp. 107-140

127

Englishman, respectively? How do the features of one’s own mother tongue – and motheraccent-

influence our perception of what stands out as contrastive or salient? But even more important are

the theoretical implications of viewing competence of a given language in terms of relatively

shared and distributed knowledge.

IV. MODELS OF PHONEMIC CATEGORY STRUCTURE The two prevailing models of phonemic category structure in Cognitive Linguistics (Mompeán,

2004: 436-444) are the radial category model and the network model. In this section we shall

briefly discuss these in relation to a usage-based cognitive phonology.

The radial category model assumes that less prototypical members are organized around a

prototypical member in terms of extensions assimilated to the category on the principle of

relative similarity:

Figure 5. The radial category model as represented in Mompeán (2004).

Gitte Kristiansen

© Servicio de Publicaciones. Universidad de Murcia. All rights reserved. IJES, vol. 6 (2), 2006, pp. 107-140

128

The resulting chaining relationships stretching out from the centre to the periphery have been

compared to the spokes of a wheel:

The specific nature of the organization has been termed a radial category, because the relationship among the members is similar to an image of spokes on a wheel. There is (or may be) a central member or members. Arranged around the central members are less central ones, which are similar to the central member, but differ from it in some respect (Nathan, 1996: 110-111)

The members `arranged together on one spoke´ do not necessarily have any kind of relationship

with members forming part of adjacent spokes:

Adjacent spokes do not necessarily have any relationship with one another, but only via a path that they can both trace back to the same center. (Nathan, 1996: 112)

Metaphorical mappings involving source domains which comprise elements such as wheels and

chains are adequate enough if lectal varieties are understood in terms of different analogical

systems which do not interact in dynamic ways with one another. However, these mappings do

not suffice if we also intend to describe the ways in which perceptual dissimilarity and intra-

phonemic contrasts among the members of phonemic categories allow for language users to

convey social differentiation through linguistic distinctiveness. Our pictorial representations,

arrows included, should also convey the possibility that contrastive relationships emerge in a

non-linear fashion. Metaphorically, then, we are in need of a more suitable source domain than

wheel. The label radial network itself is in fact more neutral in the sense that it implies a system

of interwoven relationships which need not follow a specific path.

Langacker’s (1988) network model, on the other hand, involves a category prototype,

context-induced extensions from this prototype and a schema which captures the commonality

perceived in the various extensions. The improvement with respect to the radial category model

thus lies in the fact that the network models operates with a two-level structure: the level of actual

occurrence (of often quite dissimilar variants) and a schematic unit which is an abstraction over

Towards a Usage-Based Cognitive Phonology

© Servicio de Publicaciones. Universidad de Murcia. All rights reserved. IJES, vol. 6 (2), 2006, pp. 107-140

129

usage-based events. In the case of phonemic categories, the schema corresponds to the phoneme

as such:

Figure 6. The network model as represented in Mompeán (2004).

As Mompeán explains, it is often not possible to extract one schema which represents a

generalization with respect to all the extensions (which is only natural, the very nature of a

prototype category considered). Rather, several schemas may arise which relate to different

clusters of instances:

However, it is not always possible to abstract a viable, psycholinguistically plausible schema that is fully compatible with all the members of a category. For example, not every member of the phoneme category /t/ shares the features ‘‘alveolar’’, ‘‘voiceless’’, and ‘‘stop’’, so the abstraction of a highly abstract schema which contains a feature common to all members of the category and distinguishes the category from others is impossible (Taylor, 1990). The model permits, however, the abstractions of local schemas embodying the commonality of many but not all members of the category (Bybee, 1999). Some commonality between certain members of a phoneme category may exist but the commonality may not extend to the totality of the members. One such local schema for /t/ could contain the features [voiceless], [alveolar], and [stop], shared by many but not all members of the category (Taylor, 1990). (Mompeán, 2004: 458)

Gitte Kristiansen

© Servicio de Publicaciones. Universidad de Murcia. All rights reserved. IJES, vol. 6 (2), 2006, pp. 107-140

130

The need to posit that various schemas are at work at the same time is of course related to the fact

that quite often formally very dissimilar variants are functionally operative as distinctive speech

sounds in the same lexical sets in the same language. In such cases it is not easy to draw a

coherent representation which involves only two levels of abstraction: specific language-internal

variants and one language-specific schema.

In the light of the previous discussion, it might be useful to approach the problem from a

variety of different angles. First, by applying our general knowledge about prototypicality to

phonemic categories, second, by discussing the role of lectal varieties in phonemic description

and third, by bringing in the perspectives of distributed cognition, expert analysis and folk

perception.

IV.1. Phonemic categories and prototypicality

As we have just observed, similarity in form and commonality are considered as unifying factors

in cognitive phonology. In the case of the radial network model, the relationship between the

prototype and its extensions is based on perceived similarity, and in the case of the network

model, schemas arise as generalizations embodying the commonality of their instances. In the

absence of a commonality which applies to all members, local schemas obviously provide a

category which exhibits a high degree of variation with internal cohesion. Let us observe that in a

prototype category it is only normal for family resemblances to cluster in partially overlapping

subsets, and we would certainly not expect there to be one single feature which would be

common to all the extensions (unless one opts for a classical model based on necessary and

sufficient features, or an essentialist definition). In fact, one would expect a phonemic category to

exhibit the same characteristics as any other prototype category, to varying degrees, does: (i)

absence of classical definitions, (ii) clustering of overlapping senses – or features, (iii) degrees of

representativity, and (iv) absence of clear boundaries. In the case of lexical items (cf. Geeraerts et

al., 1994: 48), the first two characteristics operate at the intensional level and the last two at the

extensional one, but combine in other ways as well: while (i) and (iv) reflect the flexibility and

vagueness that characterizes many prototype categories, (ii) and (iii) in turn result from perceived

Towards a Usage-Based Cognitive Phonology

© Servicio de Publicaciones. Universidad de Murcia. All rights reserved. IJES, vol. 6 (2), 2006, pp. 107-140

131

differences in structural weight. Nonequality and nonrigidity thus lie in the very nature of

prototypicality.

Furthermore, it is often not formal, but functional criteria which determine whether a new

subcategorization is established within a prototype category. In the case of the superordinate

category FURNITURE, for instance, ashtrays may become assimilated as (peripheral) category

members (Taylor, 1995) because they are functionally related to other such domestic artifacts

despite being dissimilar in form to other central or non-central members. In the case of phonetic

variants, from a speaker-oriented perspective the need for distinctive variants to convey social

meaning can lead to both intra-phonemic and trans-phonemic variation. From a hearer-oriented

perspective, dissimilarity would at first sight appear to be as counter-productive as lenitions are

when viewed against the principles which underlie fortitions. However, opposing tendencies need

not be incompatible. Even if a given speech sound (y) were to be classified as an instance of a

given phoneme (Y) in hearer’s own phonological system, when uttered by a speaker from a

different speech community in a context where x is expected, y could still be classified as an

instance of phoneme X, at least for the purpose of mutual understanding – as when [laɪ] is heard

and `lay´ is understood (cf. above). In other words, if “each encounter with the language leaves a

mental trace in the corpus” (Taylor, 2002: 33), and our receptive competence is experientially

grounded, the variant y will be understood to belong to category X in the speech and system of

the speaker in question – an at least ad hoc categorization based on functional, rather than

perceptual criteria.

What, then, holds a phoneme category together? Clusters based on perceptual similarity or

co-occurrence of distinct realizations in the same phonetic context, or more specifically, within

the same lexical set? Is the train of thought primarily in the direction of `as this sound is similar

to sound y it must be processed as phoneme Y´ or `as this sound occurs in a context where I

expect sound y, it should be processed as an instantiation of phoneme Y´? In the former case, the

radial category model is adequate enough, but in the latter case the network model is superior as

it enables us to work around different layers of abstraction. It is also worth noticing that one of

the implications of such a perspective is that phoneme recognition might well be lexically

Gitte Kristiansen

© Servicio de Publicaciones. Universidad de Murcia. All rights reserved. IJES, vol. 6 (2), 2006, pp. 107-140

132

mediated. If this is the case, the ideational function interacts with the social function of language

in ways which are flexible enough to allow for both of these apparently conflicting tendencies to

co-exist, to render language an efficient tool for a variety of communicative purposes at the same

time.

IV.2. Phonemic categories and lectal variation

The network model, then, allows us to work around different levels of abstraction. Would it be

possible for the model also to incorporate low-level schemas at an intermediate level of linguistic

diversity? A model capable of operating with high-level schemas, local schemas and instances

should indeed be able to reflect the taxonomic intricacies of language-internal variation in a far

more precise fashion than the radial network model. It might also be fruitful to shift the

perspective from category-internal variation (involving pure form) to a language-internal one,

reflecting form, users and functions alike. An intermediate level of `local schemas´ would

furthermore fill an important gap in a usage-based approach: that which – mediating between

parole and langue – incorporates structured variation at the level of lects. Obviously, the kind of

low-level schemas discussed in this section differ from the results of a post-hoc analysis

concerned with finding patterns in subgroups with relative similarity as the basic criterial factor.

Rather, it is an account which allows for language-internal categories to form part of the global

picture. Relative dissimilarity between (clusters of relatively similar) features and a consideration

of the ways in which these relate to their users would certainly also form part of the analysis.

The network model moreover allows us to reflect the factor of awareness as discussed in

section III.1 above. The following three-level figure draws on Tuggy’s (1993) representation of

the ambiguity-vagueness cline between lexical polysemy and homophony:

Towards a Usage-Based Cognitive Phonology

© Servicio de Publicaciones. Universidad de Murcia. All rights reserved. IJES, vol. 6 (2), 2006, pp. 107-140

133

Figure 7. Salience and awareness in phonemic and lectal categorization

In Tuggy’s analysis, the thickness and continuity of the lines iconically convey the enhanced

entrenchment, or degree of salience, of either a schema or a semantic structure, associated with

the same phonological pole. Tuggy views entrenchment in terms of enduring salience, i.e.

salience apart from relatively transitory effects such as directed attention or heightened activation

due to contextual factors. As it is precisely such transitory effects that we are interested in, the

same conventions will do for our present purposes. 7a illustrates the ideational function of

language: the high-level schema (the phoneme as a distinctive unit) receives full attention while

lectal schemas and phonetic instances are backgrounded - which does not equal saying that the

information provided is discarded.11 Priority is given to the distinctive function fulfilled by the

phoneme – an abstraction realized by specific instances. In 7b and 7c, lectal categorization

becomes increasingly more prominent. 7d and 7e both illustrate a high degree of awareness

regarding the social function of language: the high-level schema is backgrounded and attention is

on the link between linguistic form and social meaning, with an emphasis on either form, or

meaning, respectively.

IV.2. Phonemic representation and phonemic categorization: expert analysis vs. folk perception We have so far argued that regardless of whether we implement the radial category model or the

network model, our description should aim at being as refined as possible. Ideally, we would aim

Gitte Kristiansen

© Servicio de Publicaciones. Universidad de Murcia. All rights reserved. IJES, vol. 6 (2), 2006, pp. 107-140

134

at incorporating the whole range of variants in actual use in a given language - conceived of in

terms of a heterogeneous speech community and a complex social system.

If what we intend, however, is to reach a better understanding of what goes on in the mind

of the native speaker, this picture may well turn out to be rather fictitious. Even if we succeeded

in bringing all the variants in actual use under the same schematic representation, the

representation might not be very realistic. To the extent that cognition is distributed across speech

communities and cultural groups - if knowledge is only relatively shared - a realistic view cannot

posit that the individual user stores the whole range of variants in actual use within a language, be

they phonetic variants or perhaps the multiple senses of a polysemous lexeme or preposition.

Sandra and Rice (1995) convincingly question the validity of representations of vast,

global networks of senses depicted in the analysis of linguists as opposed to what ipso facto is

acquired and stored in terms of mental representations in the mind of the individual, and it is

indeed important to establish a distinction between the linguist’s attempt at providing a global

picture (one which cumulatively depicts the existence of multiple networks) and folk perception

(the relative knowledge of a more global network in the mind of the individual - and the

linguist’s attempt at reflecting more local networks). The tension involves a clash between a

perspective which in a structuralist or generativist fashion zooms in on `language structure´,

independently of the fact that there might be more than one system at work, and one which

regards language-internal variation as natural and worthy of attention. The two perspectives are

not mutually exclusive, though. Both analyses are possible and complementary - but we need to

acknowledge the differences in a clear and conscious manner.

V. CONCLUSIONS Phonetic dissimilarity and categorization have been keywords throughout the various sections.

Categorization as such is of course based on relative similarity - a cohesive factor which helps us

organize a vast amount of variation into structured sets of like components - but it also involves

the creation of subsets, established as much on the basis of relative dissimilarity. Categorization

Towards a Usage-Based Cognitive Phonology

© Servicio de Publicaciones. Universidad de Murcia. All rights reserved. IJES, vol. 6 (2), 2006, pp. 107-140

135

also applies to the social world. At more precise levels of abstractions than that of `a language´,

there are multiple social and lectal subsystems which together constitute `a society´ and `a

language´. In this respect I have distinguished between the role of hearer and speaker, and argued

that our receptive and productive competence of lectal varieties also plays a role in the

configuration of phoneme categories and inventories. In other words, a cognitive dialectology -

including a cognitive phonology - may well serve not only to mediate between `language´ and

`society´ but also to spell out in full the consequences of a truly multi-faceted approach to

phonetic variation.

I have also stressed the difference in perspective between expert analysis and folk

perception in phonemic description, and argued that the distinction is useful when theoretical

models face linguistic facts. Finally, I have examined the relative adequacy of the radial category

model and the network model and concluded that the network model seems to present a number

of advantages over the radial category model.

NOTES 1. This paper is associated with the research project HUM2005-08221-CO2-01. 2. E.g. the generativist and structuralist conceptions of language as a system (competence and langue, respectively) which is analysable independently of social and contextual factors. 3. For the importance of distributed cognition and the distinction between expert analysis and folk perception, cf. section IV.3. 4. Note that I intentionally use the term `dialect´, and not `language´: the specific nature of such low-level schemas varies considerably from one language-internal variety to another. This is one of the reasons why it is impossible for an average (adult) speaker to imitate an accent to perfection. 5. I am grateful to Raphael Berthele for introducing me to the `kiki´ and the `booba´ in his intervention on folk perception and phonosymbolism in the Theme Session Lectal Variation and the Categorization of Lectal Varieties in Cognitive Linguistics, ICLC9, Seoul. In the original experiment, Köhler (1929) called the stimuli `takete´ and `baluma´. 6. Ablaut of course serves the same differentiating purpose. In the case of English <swim, swam, swum>, the combination of three maximally distinct variants (close front i, open a, close back u) renders paradigmatic variation less ambiguous. The use of specific combinations of phonemes or morphemes in processes of derivation and declension is in this sense not entirely unmotivated. An interesting case is that of the terms <starboard and <larboard>. Both word-forms used to denote the left and right side of a ship, respectively. The terms effectively cued

Gitte Kristiansen

© Servicio de Publicaciones. Universidad de Murcia. All rights reserved. IJES, vol. 6 (2), 2006, pp. 107-140

136

both general categorization (retention of the shared element <board>) and subcategorization (addition of different pre-modifying elements). However, (or so the anectodal story goes; cf. http://en.wikipedia.org/wiki/Port_(nautical)) from an acoustic-perceptual perspective, when implemented under harsh climatic conditions at sea, the terms were not distinctive enough - and larboard was gradually replaced by <port>. 7. It is interesting to note that the processes at work are basically the same ones as in fortitions and in general mechanisms of categorization; the difference lies in the application. 8. After Labov’s (cf. Labov, 1972) initial study on Martha’s Vineyard, he left the topic of how speech features relate to social identities and social values somewhat behind to concentrate on his `attention paid to speech´ model. By drawing attention away from situational contextual factors he aimed at eliciting speaker’s vernacular, the assumption being that speech is always monitored to context. 9. When Lambert et al (1960) undertook the task of proving that we primarily evaluate speakers on the basis of their group membership rather than on the individually-based characteristics of their voice (testing in reality the existence of a group-related link between linguistic and social stereotypes), awareness also played a major role. The matched-guise technique was implemented to show that the same (bilingual) speaker was rated quite differently according to the language he or she spoke. The subjects tested thus ignored the fact that they were attributing different sets of group-related psychological attributes to one and the same person in each case and not to different individuals. A panel of judges even rated their own speech variety (French Canadian) as inferior with respect to the more prestigious variety tested (English Canadian) –a result which in all likelihood would not have been obtained if the researchers had implemented direct methods of elicitation. 10. It should be noted that Eckert’s reasoning is in line with a variety of models on style shifting opting for approaches which assign a more active role to the speaker. Instead of merely adapting himself lectally to the circumstances of a given situation, a speaker may create personae (Coupland, 2001) or engage in proactive identity construction (Walfram & Schilling-Estes, 1998).

11. It is important to note that the perspective adopted, while still category-internal, does not depict the structure of a phonemic category, but rather language-internal variation. I do not wish to argue, in this respect, that lectal varieties are schematic with respect to the features they are composed of. Just as a category such as BIRD cannot be said to be schematic with respect to features such as eggs, wings, or feathers but rather to members such as robins or penguins, a lectal category such as British English is schematic with respect to Glaswegian or Liverpudlian, but not in a direct way with respect to phonetic features: the features in question point metonymically to a lectal category. Any schematization involved (e.g. from linguistic stereotype to token) is achieved by way of such mechanisms. REFERENCES Aitchison, J. (1991). Language change: Progress or decay?. 2nd edition. Cambridge: Cambridge

University Press. Bybee, J. (1988). Morphology as lexical organization. In M.l Hammond & M. Noonan (Eds.),

Theoretical mrphology. Approaches to modern linguistics. San Diego: Academic Press, pp. 119-142.

Towards a Usage-Based Cognitive Phonology

© Servicio de Publicaciones. Universidad de Murcia. All rights reserved. IJES, vol. 6 (2), 2006, pp. 107-140

137

Bybee, J. (1999). Usage-based phonology. In M. Darnell et al. (Eds.), Functionalism and formalism in linguistics, vol. 1: General papers. Amsterdam/Philadelphia: John Benjamins, pp. 211-242.

Bybee, J. (2001). Phonology and language use. Cambridge: Cambridge University Press. Coupland, N. ( 2001). Language, situation, and the relational self: Theorizing dialect-style in

sociolinguistics. In Penelope Eckert and John. R. Rickford (eds.), Style and sociolinguistic variation. Cambridge: Cambridge University Press, pp. 185-210.

Eckert, P. (2004). The meaning of style. Online document:

http://studentorgs.utexas.edu/salsa/salsaproceedings/salsa11/SALSA11papers/eckert.pdf

Geeraerts, D. ( 1997). Diachronic prototype semantics A contribution to historical lexicology.

Oxford: Clarendon Press. Geeraerts, D. ( 2001). On measuring lexical convergence. In A. Soares da Silva (Ed.),

Linguagem e cognição. Braga: Associação Portuguesa de Linguística, pp. 51-61. Geeraerts, D. ( 2005). Lectal variation and empirical data in cognitive linguistics. In F. Ruiz

de Mendoza Ibáñez & S. Peña Cervel (Eds.), Cognitive linguistics: Internal dynamics and inter- disciplinary interaction, Cognitive Linguistics Research 32. Berlin/New York: Mouton de Gruyter, pp. 163-190.

Geeraerts, D., Grondelaers, S., & Bakema, P. (1994). The structure of lexical variation:

Meaning, naming, and context. Berlin/New York: Mouton de Gruyter. Giles, H., Mulac, A., Bradac, J. J., & Johnson, P. (1987). Speech accommodation theory: The

next decade and beyond. In M. L. McLaughlin (Ed.), Communication yearbook 10. Beverly Hills: Sage, pp. 13-48.

Halliday, M. A. K. (1978). Language as social semiotic. The social interpretation of language

and meaning. London: Edward Arnold. Holland, D., & Quinn, N. (1987). Cultural models in language and thought. Cambridge:

Cambridge University Press. Irvine, J. T., & Gal, S. (2000). Language ideology and linguistic differentiation. In P. V.

Kroskrity (Ed.), Regimes of language: Ideologies, politics, and identities. Santa Fe, NM: SAR Press, pp. 35-83.

Jakobson, R. (1960). Closing statement: Linguistics and poetics. In Thomas A. Sebeok (Ed.),

Style in language. Cambridge, Mass.: MIT Press, pp. 350-377. Kristiansen, G. (2001). Social and linguistic stereotyping: A cognitive approach to accents.

Estudios Ingleses de la Universidad Complutense, 9, 129-145.

Gitte Kristiansen

© Servicio de Publicaciones. Universidad de Murcia. All rights reserved. IJES, vol. 6 (2), 2006, pp. 107-140

138

Kristiansen, G. (2003). How to do things with allophones: Linguistic stereotypes as cognitive reference points in social cognition. In: R. Dirven et al. (Es.), Cognitive models in language and thought. Ideology, metaphors and meanings. Cognitive Linguistics Research 24. Berlin/New York: Mouton de Gruyter, pp. 69-120.

Kristiansen, G. (forthcoming a). Style-shifting and shifting styles. A socio-cognitive approach

to lectal variation. In: G. Kristiansen & R. Dirven (Eds.), Cognitive sociolinguistics. Language variation, cultural models, social systems. Berlin/New York: Mouton de Gruyter.

Kristiansen, G. (forthcoming b). Idealized cultural models: The group as a variable in the

development of cognitive schemata. In: R. M. Frank et al. (Eds.), Body, language, and mind Vol. 2: Cultural situatedness. Berlin/New York: Mouton de Gruyter.

Köhler, W. (1929). Gestalt psychology. New York: Liveright. Köhler, W. (1947). Gestalt psychology: An introduction to new concepts in modern

psychology. 2nd. Ed. New York: Liveright. Labov, W. (1972). Sociolinguistic patterns. Philadelphia: University of Pennsylvania Press. Lambert, W. E., Hodgson, R., Gardner, R. C., & Fillenbaum, S. (1960). Evaluational reactions

to spoken languages. Journal of Abnormal and Social Psychology, 60, 44-51. Langacker, R. W. (1988). A usage-based model. In Brygida Rudka-Ostyn (Ed.), Topics in

cognitive linguistics. Amsterdam/Philadelphia: John Benjamins, pp. 127-161. Langacker, R. W. (1999). Grammar and conceptualization. Berlin/New York: Mouton de

Gruyter. Lyons, J. (1977). Semantics. Cambridge: Cambridge University Press. Mompeán, J. A. (2004). Category overlap and neutralization: The importance of speakers’

classifications in phonology. Cognitive Linguistics, 15(4), 429-469. Nathan, G. S. (1986). Phonemes as mental categories. Proceedings of the Berkeley Linguistic

Society, 12, 212-223. Nathan, G. S. (1994). How the phoneme inventory gets its shape: Cognitive grammar’s view

of phonological systems. Rivista di Linguistica, 6, 275-287. Nathan, G. S. (1996). Steps towards a cognitive phonology. In B. Hurch & R. Rhodes (Eds.),

Natural phonology: The state of the art. Berlin/New York: Mouton de Gruyter, pp. 107-120.

Nathan, G. S. (1999). What functionalists can learn from formalists in phonology. In M.

Darnell et al. (Eds.), Functionalism and formalism in linguistics, vol. 1: General papers. Amsterdam/Philadelphia: John Benjamins, pp. 305-327.

Towards a Usage-Based Cognitive Phonology

© Servicio de Publicaciones. Universidad de Murcia. All rights reserved. IJES, vol. 6 (2), 2006, pp. 107-140

139

Niedzielski, N., & Preston, D. (2000). Folk linguistics. Berlin/New York: Mouton de Gruyter. Nunberg, G. (1978). The Pragmatics of reference. Bloomington, Indiana: Indiana University

Linguistics Club. Nuyts, J. (2005). Brothers in arms? On the relations between Cognitive and Functional

Linguistics. In F. J. Ruiz de Mendoza Ibáñez & M. S. Peña Cervel (Eds.), Cognitive linguistics. Internal dynamics and interdisciplinary interaction, Cognitive Linguistics Research 32. Berlin/New York: Mouton de Gruyter, pp. 69-100.

Orton, H., Sanderson, S., & Widdowson, J. (Eds.) (1978). The linguistic atlas of England.

London: Croom Helm. Purnell, T., Idsardi, W. J., & Baugh, J. (1999). Perceptual and phonetic experiments on

American English dialect identification. Journal of Language and Social Psychology, 18, 10-30.

Sandra, D., & Rice, S. (1995). Network analyses of prepositional meaning: Mirroring whose

mind – the linguist’s or the language user’s?. Cognitive Linguistics, 6(1), 89-130. Sharifian, F. (2003). On cultural conceptualisations. Journal of Cognition and Culture. 3(3),

187-206. Stampe, D. (1979). A dissertation on natural phonology. New York: Garland Publishing. Tajfel, H., & Turner, J. C. (1979). An integrative theory of intergroup conflict. In W. G.

Austin & S. Worchel (Eds.), The social psychology of intergroup relations. Monterey, CA: Brooks/Cole, pp. 33-47.

Taylor, J. R. (1990). Schemas, prototypes, and models: In search of the unity of the sign. In S.

L. Tsohatzidis (ed.), Meanings and prototypes: Studies in linguistic categorization. London: Routledge, pp. 521-534.

Taylor, J. R. (1995). Linguistic categorization. Prototypes in linguistic theory. 2nd. edition.

Oxford: Clarendon Press. Taylor, J. R. (2002). Cognitive grammar. Oxford: Oxford University Press. Tuggy, D. ( 1993). Ambiguity, polysemy, and vagueness. Cognitive Linguistics, 4(3), 273-290. van Bezooijen, R., & Gooskens, Ch. (1999). Identification of language varieties. The

contribution of different linguistic levels. Journal of Language and Social Psychology, 18(1), 31-48.

Werner, H. (1934). L’unité des sens. Journal de Psychologie Normale et Pathologique, 31,

190-205.

Gitte Kristiansen

© Servicio de Publicaciones. Universidad de Murcia. All rights reserved. IJES, vol. 6 (2), 2006, pp. 107-140

140

Werner, H. (1957). Comparative psychology of mental development. New York: International Universities Press.

Werner, H., & Wapner, S. (1952). Toward a general theory of perception. Psychological

Review, 59: 324-38. Wolfram, W., & Schilling-Estes, N. ( 1998) American English: Dialects and variation.

Oxford: Blackwell.